A breakthrough in radiology and digital health may soon change how clinicians assess one of the most pervasive health risks of our time: chronic stress. Using advanced deep learning techniques, a research team from Johns Hopkins University has identified the first imaging biomarker capable of capturing the biological footprint of long-term stress, directly from routine CT scans.
The findings, presented at the annual meeting of the Radiological Society of North America (RSNA), highlight the growing potential of AI to extract new, actionable insights from existing medical data.
Chronic stress affects millions worldwide and is linked to a wide spectrum of physical and mental health problems, including hypertension, depression, insomnia, immune dysfunction and cardiovascular disease. Yet despite its impact, healthcare has lacked an objective and scalable method to quantify the cumulative burden of stress. Traditional indicators, such as cortisol levels or self-reported questionnaires, offer only fragmented or momentary snapshots. This makes early identification and risk stratification difficult, particularly in everyday clinical settings.
Deep learning model
To bridge this gap, lead researcher Dr. Elena Ghotbi, postdoctoral fellow at the Johns Hopkins University School of Medicine, developed a deep learning model capable of automatically measuring adrenal gland volume on standard chest CT scans. The adrenal glands play a central role in the body’s stress response. Unlike single cortisol measurements, adrenal volume reflects chronic activation of the stress system, functioning as a biological barometer of long-term overload.
The research team validated this AI model using data from 2,842 participants in the Multi-Ethnic Study of Atherosclerosis (MESA). This unique dataset combines CT imaging, salivary cortisol collected multiple times per day, detailed stress questionnaires and a broad panel of allostatic load markers, representing the cumulative physiological wear and tear caused by stress. Such an integrated dataset is extremely rare; according to the researchers, it is likely the only cohort suitable for developing and validating an imaging-based stress biomarker.
Adrenal Volume Index
From each CT scan, the model calculated an Adrenal Volume Index (AVI), defined as adrenal volume relative to patient height. Statistical analyses showed clear associations between AVI and multiple validated indicators of chronic stress. Higher AVI scores correlated with elevated cortisol, greater allostatic load, higher perceived stress and depressive symptoms. Importantly, AVI also predicted long-term clinical outcomes: each incremental rise in AVI was linked to increased risk of heart failure and overall mortality during a follow-up period of up to ten years.
Senior author Dr. Shadpour Demehri emphasizes the practical value of this innovation. Because chest CT scans are already performed millions of times per year for a wide range of clinical indications, the biomarker can be implemented without adding new tests, procedures or radiation exposure. “For the first time, we can visualise the long-term burden of stress using a scan patients already receive in hospitals across the country,” he noted.
Co-author and stress-research pioneer Prof. Teresa Seeman adds that the study represents a major step forward in operationalising the cumulative impact of stress on the human body. An insight that until now relied mainly on behavioural and biochemical assessments.
With the introduction of this AI-driven imaging biomarker, clinicians may soon gain a powerful new tool for early detection, personalised risk stratification and targeted prevention of stress-related diseases. The research opens the door to applying the biomarker across a range of conditions in middle-aged and older adults, potentially transforming how healthcare recognises and addresses the silent toll of chronic stress.
Stress diagnosis within minutes
Earlier this year we wrote about a portable ‘lab-on-a-chip’ saliva test, developed by researchers at the University of Cincinnati. This test measures stress hormones such as cortisol and DHEA through saliva, offering a fast, non-invasive point-of-care test for early detection of stress-related conditions like depression and anxiety. Patients place a saliva sample into a small device, which delivers results to a smartphone within minutes, dramatically reducing diagnostics that normally take days.
By combining psychometric assessments with objective biomarker data, the system can identify individuals who may be overlooked due to stigma or limited symptom awareness. The technology also shows promise beyond mental health: the same chip can measure troponin from a drop of blood for cardiac monitoring. With further clinical validation, this innovation could support hybrid care models and enable faster, more personalised healthcare.
